Commit Graph

4 Commits

Author SHA1 Message Date
Wing Lian
05d19d2037 remove debugging, use gpt2 since starcoder requires consent
Some checks failed
pre-commit / pre-commit (push) Has been cancelled
PyTest / test (3.10) (push) Has been cancelled
PyTest / test (3.9) (push) Has been cancelled
2023-06-13 21:32:47 -04:00
Wing Lian
61f44f311e fix packing for tokenizers that don't use a bos_token when the bos token and eos token are both the same 2023-06-13 21:26:13 -04:00
Wing Lian
0136f510f2 don't worry about duplicate code here 2023-05-31 12:05:43 -04:00
Wing Lian
9b8585dc70 fix packing so that concatenated sequences reset the attention 2023-05-31 11:38:52 -04:00